DeepSomatic ONT case study

March 5, 2026 · View on GitHub

In this case study, we show an example of running DeepSomatic on ONT data. We use HCC1395 as an example for this case study.

Data details

For this case-study, we use HCC1395 as an example. We run the analysis on chr1 that we hold out during training.

Please see the metrics page for details on runtime and data.

Prepare environment

Tools

Docker will be used to run DeepSomatic and hap.py,

Download input data

We will be using GRCh38 for this case study.

BASE="${HOME}/deepsomatic-ont-case-study"

# Set up input and output directory data
INPUT_DIR="${BASE}/input/data"
OUTPUT_DIR="${BASE}/output"

## Create local directory structure
mkdir -p "${INPUT_DIR}"
mkdir -p "${OUTPUT_DIR}"
mkdir -p "${OUTPUT_DIR}/sompy_output"

# Download bam files to input directory
HTTPDIR=https://storage.googleapis.com/deepvariant/deepsomatic-case-studies/deepsomatic-chr1-case-studies
# Download the reference files
curl ${HTTPDIR}/GCA_000001405.15_GRCh38_no_alt_analysis_set.chr1.fna > ${INPUT_DIR}/GCA_000001405.15_GRCh38_no_alt_analysis_set.chr1.fna
curl ${HTTPDIR}/GCA_000001405.15_GRCh38_no_alt_analysis_set.chr1.fna.fai > ${INPUT_DIR}/GCA_000001405.15_GRCh38_no_alt_analysis_set.chr1.fna.fai

# Download the bam files
curl ${HTTPDIR}/HCC1395_ont.normal.chr1.bam > ${INPUT_DIR}/HCC1395_ont.normal.chr1.bam
curl ${HTTPDIR}/HCC1395_ont.normal.chr1.bam.bai > ${INPUT_DIR}/HCC1395_ont.normal.chr1.bam.bai
curl ${HTTPDIR}/HCC1395_ont.tumor.chr1.bam > ${INPUT_DIR}/HCC1395_ont.tumor.chr1.bam
curl ${HTTPDIR}/HCC1395_ont.tumor.chr1.bam.bai > ${INPUT_DIR}/HCC1395_ont.tumor.chr1.bam.bai

# Download truth VCF
DATA_HTTP_DIR=https://storage.googleapis.com/deepvariant/deepsomatic-case-studies/SEQC2-S1395-truth
wget -P ${INPUT_DIR} "${DATA_HTTP_DIR}"/High-Confidence_Regions_v1.2.bed
wget -P ${INPUT_DIR} "${DATA_HTTP_DIR}"/high-confidence_sINDEL_sSNV_in_HC_regions_v1.2.1.merged.vcf.gz
wget -P ${INPUT_DIR} "${DATA_HTTP_DIR}"/high-confidence_sINDEL_sSNV_in_HC_regions_v1.2.1.merged.vcf.gz.tbi

Running DeepSomatic with one command

DeepVariant pipeline consists of 3 steps: make_examples_somatic, call_variants, and postprocess_variants. You can run DeepSomatic with one command using the run_deepvariant script.

Running on a CPU-only machine

BIN_VERSION="1.10.0"

sudo docker pull google/deepsomatic:"${BIN_VERSION}"

sudo docker run \
-v ${INPUT_DIR}:${INPUT_DIR} \
-v ${OUTPUT_DIR}:${OUTPUT_DIR} \
google/deepsomatic:"${BIN_VERSION}" \
run_deepsomatic \
--model_type=ONT \
--ref=${INPUT_DIR}/GCA_000001405.15_GRCh38_no_alt_analysis_set.chr1.fna \
--reads_normal=${INPUT_DIR}/HCC1395_ont.normal.chr1.bam \
--reads_tumor=${INPUT_DIR}/HCC1395_ont.tumor.chr1.bam \
--output_vcf=${OUTPUT_DIR}/HCC1395_deepsomatic_output.vcf.gz \
--sample_name_tumor="HCC1395Tumor" \
--sample_name_normal="HCC1395Normal" \
--num_shards=$(nproc) \
--logging_dir=${OUTPUT_DIR}/logs \
--intermediate_results_dir=${OUTPUT_DIR}/intermediate_results_dir \
--regions=chr1

NOTE: If you want to run each of the steps separately, add --dry_run=true to the command above to figure out what flags you need in each step. Based on the different model types, different flags are needed in the make_examples step.

--intermediate_results_dir flag is optional. By specifying it, the intermediate outputs of make_examples_somatic and call_variants stages can be found in the directory.

sudo docker pull pkrusche/hap.py:latest
# Run hap.py
sudo docker run \
-v ${INPUT_DIR}:${INPUT_DIR} -v ${OUTPUT_DIR}:${OUTPUT_DIR} \
pkrusche/hap.py:latest \
/opt/hap.py/bin/som.py \
-N ${INPUT_DIR}/high-confidence_sINDEL_sSNV_in_HC_regions_v1.2.1.merged.vcf.gz \
${OUTPUT_DIR}/HCC1395_deepsomatic_output.vcf.gz \
-r ${INPUT_DIR}/GCA_000001405.15_GRCh38_no_alt_analysis_set.chr1.fna \
-o ${OUTPUT_DIR}/sompy_output/deepsomatic.chr1.sompy.output \
--feature-table generic \
-R ${INPUT_DIR}/High-Confidence_Regions_v1.2.bed \
-l chr1

The output:

      type  total.truth  total.query    tp  fp   fn  unk  ambi    recall  recall_lower  recall_upper   recall2  precision  precision_lower  precision_upper  na  ambiguous  fp.region.size   fp.rate
0   indels          133          103    87  16   46    0     0  0.654135      0.570673      0.730971  0.654135   0.844660         0.765604         0.904679   0          0       248956422  0.064268
1     SNVs         3440         2663  2617  46  823    0     0  0.760756      0.746284      0.774789  0.760756   0.982726         0.977240         0.987165   0          0       248956422  0.184771
5  records         3573         2766  2704  62  869    0     0  0.756787      0.742516      0.770643  0.756787   0.977585         0.971558         0.982614   0          0       248956422  0.249040